What is allowed on this site
A review site is only worth reading if you can check it. Here is the rule set we apply to every product.
Read the rules-
Which page we read
We start with the manufacturer page, not the review page
Specifications come from the product page the brand controls, or from a retailer listing when the brand does not publish a spec sheet. We record the exact URL and the date we read it. That is the only way a number on this site can still be checked a year from now.
Written down so you can hold us to it, and so you can reproduce our work.
-
What we leave out
A value we cannot source does not go on the page
If a source does not publish a waste bin volume, we show a dash. We do not estimate it from a photograph, we do not average it across models, and we do not fill the gap with a number that reads plausibly. An empty field is more useful than a confident guess.
-
Prices and their dates
Prices are dated observations, never live quotes
Retail prices move and sellers disagree. Every price on this site carries the source URL and the date we read it. Two retailers listed the same PetSafe model at $399.99 and $171.00 on the same day, and we publish the range rather than picking the flattering number.
-
How complaints are counted
Complaint records are quoted, not tallied by us
We publish owner complaint patterns that appear in the cited sources. We do not scrape, score or aggregate reviews ourselves, and we do not convert them into a star number, because a star number implies a measurement we did not perform.
-
Scores we do not give
We publish no rating of our own yet
Our bench is not running. When it is, first-party measurements will appear here, clearly marked as ours, dated, and published whether or not they flatter the products we earn from. Until then the honest thing to publish is the specification and the complaint record.
-
How the score is worked out
The setup-risk score is arithmetic, not instinct
Five published facts each add one point: 2.4 GHz only, a one-year warranty, a waste bin under 8 litres, measured noise at 40 dB or above, and a warranty we could not confirm. Zero to one point is lower risk, two to three is moderate, four or more is higher. You can recompute it from the table. The score is written for hardware that runs itself, so it appears on litter boxes and feeders. Water fountains and indoor potties fail in different ways, and those records carry the running cost instead: how often the maker tells you to replace a filter or a pad roll.
-
What commission changes
Money never reaches the ranking
Affiliate commission changes which products we can afford to link, not which products we list or how we order them. We do not accept payment for placement, we do not gate reviews behind a programme, and no product is badged as a pick because of what it pays.
-
How we correct errors
Corrections are published, not quietly patched
When a source corrects itself or a specification changes, the page updates and the observation date moves with it. If we got something wrong, tell us and we will fix it and say so.
What each record contains
The dataset is one JSON document with a stable schema.
Every product carries the same 34 to 37 fields, including the ones with a null value, so a consumer can tell "not published" apart from "zero".
The record also holds the animals the maker sells the product for, read from the page we cite. A feeder sold for cats and small dogs is listed for both. A litter box with a raised entry to keep dogs out is listed for cats only. That is a fact about the product, not about the reader: the same record serves both, and no page on this site changes its content because of where you are reading it from.
| Field group | Example fields | Rule |
|---|---|---|
| Identity | id, cluster, species, brand, model, url | The id is stable and is the join key across the API, the MCP tools and the page URLs. Species lists the animals the maker sells the product for. |
| Family | cluster, servesNote, runningCost | Cluster is the product family: smart-litter-box, automatic-feeder, water-fountain or indoor-potty. Running cost is the published consumable cadence, shown instead of a risk score where the score does not apply. |
| Price | priceUsd, priceLowUsd, priceHighUsd, priceSource, priceObservedOn | Always a dated observation at a named source. Null when no dated price was found. |
| Specification | wasteBinL, interiorL, catWeightMinLb, entryHeightIn, noiseDb, wifiBands, warrantyMonths | Null means the source did not publish it. Never inferred. |
| Provenance | sourceUrl, sourceLabel, observedOn, priceVerification, noiseVerification | Per-field verification where a value can come from more than one kind of source. |
| Editorial | strengths, tradeoffs, documentedComplaints, recurringCostNote | Derived from cited sources. Each entry is a quotation or a direct restatement, not an opinion. |
Use it yourself
The whole dataset is one request away.
It is described in an OpenAPI document so you can generate a client from it.
/openapi.json
The machine-readable description of the endpoints, parameters and error states.
/.well-known/ai-catalog.json
The resources this site publishes, in the discovery format agents read.
Corrections
Something wrong here? If a figure on this site is out of date, out of context or simply wrong, tell us. We will correct it and update the observation date on the page. Corrections are not quietly patched and they are not hidden.
Check one figure against the rule
Take any product, pick one published number and check it against the source it came from. That is the whole method.
Open the finder